Whole industries rarely disappear because of AI, but specific business models do. Exposure is highest where a business sells access to information, routine expertise or effort billed by the hour, and lowest where it owns scarce assets, trusted relationships or regulated accountability. Boards should test their own revenue lines, not their sector’s headlines.
AI investors and technology leaders now ask openly which industries survive AI. This article is AIPEX’s own answer to the version of that question a board can act on: how exposed is our business model, and how would we know?
Why is “which industries survive” the wrong question?
Sector labels are too broad to guide a decision. Within professional services, legal, accounting or consulting, some firms sell scarce judgement and trusted accountability, while others sell hours of routine work. Within retail, some sell convenience that AI can improve, and others sell product knowledge that AI can give away. The same technology can strengthen one business and hollow out its neighbour.
The useful unit of analysis is the revenue line. What exactly does the customer pay for, and could AI give them that more cheaply, from someone else, or without a supplier at all?
Which business models are most exposed?
Five features raise exposure. None is decisive on its own, but a revenue line with three or more deserves the board’s attention.
- Pricing by effort. If the price reflects hours or headcount, and AI cuts the hours, customers will expect the price to fall.
- Selling information that is becoming abundant. Where value lies in knowing or finding something, AI narrows the gap between expert and customer.
- Routine, rules-based work. Work that follows clear steps with checkable outputs is the easiest for AI to perform.
- Standing between two parties. Intermediaries that match, route or translate are exposed when AI lets the parties deal directly.
- Weak switching costs. If a customer can move easily, a cheaper AI-enabled rival will not have to be much better to win.
The mirror image shows what protects a business model. Businesses are more resilient where they own physical assets, networks or licences that are hard to replicate; hold accountability that customers or regulators insist a named organisation carries; have relationships built on trust over years; or own data about customers and operations that others cannot obtain. These do not make a business immune. They mean AI is more likely to lower its costs than to remove its customers. The protections can also be built deliberately: deeper integration with customers, services that combine AI with accountable human judgement, or data assets that grow with every transaction. A board that sees exposure rising should ask which of these protections the business could strengthen, and at what cost, before a competitor forces the question.
How should a board test its own exposure?
A simple exercise, run with finance, turns the debate into numbers. For each significant revenue line, score the five exposure features and the four protections, then look at how much revenue sits where exposure is high and protection is low.
| Exposure | Protection | What it usually means | Board response |
|---|---|---|---|
| High | Low | Revenue at risk of price pressure or loss | Reprice, redesign the offer or plan an orderly decline |
| High | High | Pressure on margin, but the customer stays | Use AI to cut cost before competitors force the price down |
| Low | Low | Stable for now, but no moat | Watch closely; invest in a protection |
| Low | High | Resilient | Use AI to extend the advantage into adjacent revenue |
The first row is where boards most often underreact, because the revenue still looks healthy today. The warning signs usually arrive first as pressure in renewals and tenders, not in the annual results.
Why should the board use its own benchmarks?
Public AI benchmarks measure what models can do on standard tests. They say little about whether AI can do your customers’ work to the standard they pay for. A board that relies on headlines will either panic too early or relax too long.
A better approach is to build internal benchmarks: take a sample of real work from each exposed revenue line and test, under controlled conditions, how well current AI tools perform it and at what cost. Repeat the test every six months. The trend matters more than any single result, and it gives the board evidence instead of opinion. The finance function has a central role here, as our article on the CFO’s AI transformation mandate explains.
What should the board ask management?
- What share of our revenue is priced by effort, and what happens to it if AI halves the effort?
- Which of our revenue lines sell information or routine expertise that AI is making abundant?
- Where would a customer be able to do for themselves, with AI, what they pay us for today?
- What protects our largest revenue lines, and is that protection growing or eroding?
- Have we tested AI on a sample of our own work, and when will we test again?
Where to start
Our briefing on how AI creates competitive advantage and the questions every board should ask looks at the other side of exposure: where AI can strengthen a position rather than erode it. For boards that suspect competitors are moving faster but not why, the step on its four-step path that fits is step 2, Assess: the AI Readiness Assessment (£999 + VAT) scores 30 areas across six domains, with risk flags, a prioritised roadmap and a 60-minute debrief. The AI Competitive Advantage Scorecard below gives a quicker first read.
Frequently asked questions
Which business models are most exposed to AI?
Those that price by effort, sell information that is becoming abundant, perform routine rules-based work, sit between two parties, or have weak switching costs. A revenue line with three or more of these features deserves close board attention.
What protects a business from AI disruption?
Scarce physical assets, networks or licences; accountability that customers or regulators insist a named organisation carries; relationships built on long-term trust; and data others cannot obtain. These make it more likely AI lowers the business's costs than removes its customers.
Why should a board build its own AI benchmarks?
Public benchmarks test models on standard tasks, not on your customers' work. Testing current AI on a sample of real work every six months shows the trend in your own exposure. See The CFO's AI Mandate →